Motto : Crapula ingenium offuscat. Traduction : "le bec du perroquet qu'il essuie, quoiqu'il soit net" (Pascal).

Ce blog est ouvert pour faire connaître les activités d'un groupe de recherches, le Séminaire de métaphysique d'Aix en Provence (ou SEMa). Créé fin 2004, ce séminaire est un lieu d'échanges et de propositions. Accueilli par l'IHP (EA 3276) à l'Université d'Aix Marseille (AMU), il est animé par Jean-Maurice Monnoyer, bien que ce blog lui-même ait été mis en place par ses étudiants le 4 mai 2013.

Mots-clefs : Métaphysique analytique, Histoire de la philosophie classique, moderne et contemporaine,

Métaphysique de la cognition et de la perception. Méta-esthétique.

Austrian philosophy. Philosophie du réalisme scientifique.

vendredi 27 octobre 2017

[ Une première journée de métaethique s'est tenue à Gordes (Vaucluse) le samedi 5 novembre 2016, autour de John SKORUPSKI et de son grand ouvrage The Domain of Reasons (Oxford University Press). Sont également intervenus lors de cette journée : Davide Fassio, Pascal Engel et Isabelle Pariente-Butterlin. Une publication en français est programmée de cette rencontre sous l'intitulé : Raisons. D'autres interventions seront progressivement communiquées. Dans l'attente, le SEMa est heureux de pouvoir proposer dans son blog une version remaniée de l'intervention du Professeur Skorupski en version originale. ]

Scanlon on the Ontology of Reason Relations v.2

John Skorupski

§1 T. M. Scanlon’s book, Being Realistic about Reasons,[1] is a general account of reasons and normativity which has, in particular in chapter 2, a notable account of what it is for normative objects to exist. It is this account that I would like to discuss here.
In his introductory lecture, Scanlon notes two ways in which the meta-ethical debate has changed. Up to the later 1970s, when John Mackie published Ethics: Inventing Right and Wrong, the focus was very largely on morality. Since then it has broadened to the general notion of normativity:

although morality is still much discussed, a significant part of the debate concerns practical reasoning and normativity more generally: reasons for action, and, even more broadly, reasons for belief and other attitudes, which are increasingly recognised as normative …[2]

The second difference, he says, is that the focus then was on motivation, especially the question of how moral judgements could motivate (if they did), whereas now, though the importance of questions about motivation is not disputed, the focus has moved to the idea of a normative reason – for action, belief or “other attitudes”. Scanlon rather modestly remarks that this claim about meta-ethics as a whole may be tendentious, since the two shifts may be shifts only in his own thinking. However I think most people interested in the subject would agree that these two shifts in the debate have indeed been quite general, if by no means universal. The broadening of attention has had various sources, not least the natural development of the subject; but Scanlon’s thinking has unquestionably been a very important influence on it. We now increasingly think broadly in terms of meta-normativity, rather than narrowly in terms of meta-ethics alone, and the idea that the concept of a reason is the key normative concept is very much in the air.
Scanlon’s lectures reflect both trends. He describes himself as a Reasons Fundamentalist. This is the name he gives to the view “that truths about reasons are not reducible to or identifiable with non-normative truths, such as truths about the natural world of physical objects, causes and effects, nor can they be explained in terms of notions of rationality or rational agency that are not themselves claims about reasons.”[3] He notes that reasons might be fundamental in a further sense, namely, that other normative concepts might all be analysable in terms of the notion of a reason. This is a view to which he says he is inclined, but recognises as controversial and does not propose to defend in the book (although he later adopts it as a “working hypothesis”[4]). 
I agree with Scanlon’s Reasons Fundamentalism. I also find the hypothesis that all normative concepts are analysable in terms of reason relations highly plausible.[5] True, we need more to establish that the concept of a reason relation is the fundamental normative concept. What is needed is the negative thesis that the circle of normative concepts cannot be analysed equally well in terms of some normative concept other than that of a reason relation, for example the concept good. For if something like that could be done there would turn out to be more than one way of analysing some normative concepts in terms of others; none of them would be describable, therefore, as fundamental. I am myself sceptical as to whether there is any analysis of normativity that is as good as or better than the one in terms of reasons. However we do not need to pursue the matter here. Scanlon’s ontological discussion – which he in effect takes to be a discussion of the normative domain in general – is conducted in terms of reason relations, and I shall follow him in that. But the analysis would work, if it works, whatever one took fundamental normative properties to be.
The main aim of Scanlon’s discussion is to offer “a qualified defence of a realistic cognitivism about reasons.”[6] By ‘cognitivism’ Scanlon understands the familiar view that propositions about reasons can be true or false and, as I believe he would accept, true or false in a perfectly standard, generic sense of ‘true’ and ‘false’ – not one that is special to the normative case. ‘Realistic’ he uses in a more idiosyncratic way.[7] His version of cognitivism is realistic in the sense that it accepts that some normative claims may not have determinate truth values – this against what Scanlon would presumably take to be the unrealistic insistence that every normative claim must have a determinate truth value. I think he means that such insistence would be unrealistic in a familiar non-philosophical sense of the word. Clearly then, being ‘realistic’ in this sense does not entail endorsement of normative realism, in any of the common philosophical senses in which realism about the normative is opposed to anti-realism, irrealism, or quasi-realism. On the contrary, it seems to me that Scanlon rejects those forms of realism, while also rejecting anti-realism, irrealism or quasi-realism. As I understand him he rejects both wings of this debate as inappropriately metaphysical. Let us pursue this a little further.

§2 Like several other writers on normativity, Scanlon takes reasons to be facts which stand in a reason relation to a person and a response-type by that person. He further distinguishes between the fact or facts that constitute a reason and the circumstances in virtue of which they constitute that reason. (So one might say, for example, that the fact that the building is on fire is a reason for you to enter – given the circumstance that you are a fireman who has been put in charge of saving as much as possible of burning buildings.) So

“is a reason for” is a four-place relation, R(p, x, c, a) holding between a fact, p, an agent x, a set of conditions c, and an action or attitude a. This is the relation that holds just in case p is a reason for a person x in situation c to do or hold a.[8]

On this account, as he notes, reasons, that is, the facts that are reasons, do not pose any special ontological question that the notion of a fact in general does not pose. As to this, Scanlon understands by facts “the reflection of true thoughts;” he distances himself from a correspondence notion of a fact as a truth-maker for a proposition.[9] His notion of fact is effectively the Fregean notion. Reasons are facts in that special sense of ‘fact’ in which facts are truths, as against what ‘makes’ propositions true.
This still leaves open the question of the ontological status of reason relations, as against reasons. Scanlon’s response in Lecture 2, though not wholly new, is developed in an innovative way. The basic idea is what he calls the “domain-centered view”[10] of existence. According to this, questions about what exists should be thought of as internal to domains of discourse such as “mathematics, science, and moral and practical reasoning.”[11] There are

no general, domain-independent conditions of “existence” such that the various existential claims made in every domain entail or presuppose that entities of the kind they refer to fulfil these conditions … the only thing common to existential claims across all domains is the purely formal logic of the existential quantifier … the conditions required for objects in different domains to exist varies [sic] from domain to domain.[12]

On the whole, then, and despite the elasticity of the term ‘realism’ in current meta-normative debates, it would seem wrong to classify Scanlon as a normative realist. That term, I believe, is most often used to denote the view not just that normative claims express propositions, but that those propositions are true only if they have truth-makers that make them true, and finally, that some of them are true. Let’s drop the last thesis, since from the meta-normative point of view it adds a claim that should be treated separately. Realism about the normative, as I will use the term here, is simply the thesis that normative propositions fall under a correspondence conception of truth – leaving open the question whether any of them are true. Typical error theorists about the normative are in this sense realists: they hold that normative claims do indeed fall under a correspondence conception of truth, but they think they are never true, because there are no facts that make them true. In contrast Scanlon is not a realist in this sense; as just noted, he explicitly denies that normative truths must have truth-makers. He is a cognitivist, but he is not a realist.

§3 Scanlon also distances his view from expressivism.  Here however the point of difference is left somewhat obscure. He notes that expressivists like Allan Gibbard and Simon Blackburn say that their position allows them to accept that there are normative truths, and therefore also that there are normative facts, in the Fregean sense of fact. And of course they too will agree that these normative truths do not have truth makers.
It is not in fact possible, I believe, to establish a difference between Scanlon’s cognitivism and Gibbard’s complex expressivism or Blackburn’s ‘quasi-realism’ without directly challenging the expressivists’ right to lay claim to a (standard, generic) notion of truth for the case of normative propositions – or indeed their right to lay claim to the notion of a normative proposition. This is something Scanlon does not attempt, perhaps because he seeks agreement rather than disagreement. As a result, however, his rejection of expressivism looks a bit arbitrary.
So let’s consider the issue a little more closely. Take, first, the claim that an utterance of the sentence ‘pleasure is good’ in some contexts C constitutes a speech act of saying that pleasure is good. We can assume that the expressivist will have an account of when it is appropriate to perform that speech act. His next claim is that utterance of the sentence ‘It is true that pleasure is good’, in the same contexts C, is a performance of the same speech act. Call this the claim of speech act identity. Of course there are obvious ways in which the two speech acts are not exactly the same – different words are used. But let’s grant, for the sake of the argument, that we have some understanding of how this claim about speech act identity is intended.
Contrast now the claim that the proposition that pleasure is good is the same proposition as the proposition that it is true that pleasure is good. Call this claimed identity the propositional identity. It rests on a view of truth which one might call the simple view, or the ‘no theory’ view, or (with reservations) the minimalist view. The simple view will have something to say about occurrences of ‘is true’ as a predicate, as well as occurrences of ‘It is true that’ as an operator, but we don’t need to pursue that here. At any rate some cognitivists, of whom I’m one, agree with this simple view of truth. How then do these cognitivists differ from expressivists? The difference is in the order of explanation. For the simple cognitivist, the claim about speech act identity, in the sense in which it is true, is true in virtue of the propositional identity. The speech acts are identical in the sense that they have the same content, that is, they are assertions of the same proposition. That explanation cannot work for the expressivist, however, since the whole point about expressivism with respect to such utterances as ‘pleasure is good’ is that, whether or not they can innocuously, at the end of the analysis, be said to be quasi-assertions of quasi-propositions, which may be quasi-true or quasi-false, they cannot be said to be assertions of propositions in some antecedent sense of proposition that an expressivist can take for granted as applying to normative discourse ab initio. To accept, ab initio, that normative claims are assertions of propositions would remove any space for a distinctive expressivist semantics for normative sentences.
The cognitivist relies on the notion of same proposition in understanding the relevant notion of same speech act; he can further argue that in understanding the relevant notion of speech act identity we implicitly rely on a grasp of the proposition that the speech-act asserts, and on the simple theory of truth as applied to propositions. The expressivist will have to do it the other way round. He will have to appeal to the speech act identity as a way of constructing propositional, or rather quasi-propositional, identity, and quasi-propositional truth, for the normative case. Whether this can be done, without at some stage or other begging the question by implicitly appealing to an already understood notion of propositional identity, is moot.
In contrast, cognitivists take it as evident that normative assertions are assertions like any others.  They can be happy with truth-conditional semantics, or any form of semantics for normative discourse that (i) is uniformly applied to assertoric discourse in general, as well as normative discourse in particular, and (ii) captures the basic point that assertion is assertion of judgeable content (to use the Fregean expression). So long as these two points are respected, semantics is not, for the cognitivist, a main issue.
Let me call this standpoint, combined with minimalism about truth, simple cognitivism. Before proceeding, I should note the common view that minimalism about truth is not compatible with a truth-conditional account of meaning. If that view were correct, a simple cognitivist would have to give an account of meaning that was not truth-conditional. Perhaps it would be a use theory. Whatever the theory was, simple cognitivism would require it to preserve the point that normative assertions express judgeable contents. However I do not myself agree that this common view is correct; it seems to me to ascribe an over-ambitious role to truth-conditional semantics. Such semantics should not be seen, in my view, as being in the business of elucidating philosophically what meaning, or correlatively, understanding is. You can engage in truth-conditional semantics for a particular language without having a philosophical view of the nature of meaning and understanding. If your semantics works, it gives a systematic account of a particular language which has the property that knowing it, together with knowing that it is true in virtue solely of the conventions of the language in question, suffices to understand the language.
To be sure, there is also another question, and a good one, about whether what Michael Dummett called a ‘full-blooded’ account of understanding, or grasp of concepts, should rely on the notion of truth, or whether it should instead rely on characterising concepts in terms of norms for introducing and eliminating them in thought. It may be that minimalism about truth leads cognitivism to the latter option. But my present point is that that, if true, in no way prevents a simple cognitivist from talking straightforwardly about the truth conditions for normative sentences. The no-theory view of truth is not the view that there is no such thing as truth.

§4 Since Scanlon does not discuss these issues about truth and meaning in his lectures, I am not sure that he can be described as a simple cognitivist. On the other hand I’m fairly sure that that’s what he should be.  This position is distinct from expressivism, but at the same time is not realist in the meta-normative sense previously described. And Scanlon is neither an expressivist nor a realist in that sense.
It would be a further step to describe him as an irrealist, however. Now in view of what I’ve said so far, this may look like a merely terminological question. Perhaps it is. However I think we need to probe further before deciding whether or not it is, because the issues are quite elusive.
At the beginning of Lecture 2, which deals with “metaphysical” objections to his position, Scanlon states what he says it is “natural to describe … as an ontological objection” to his form of cognitivism about reasons, namely
that the idea that there are irreducible normative truths has implications that are incompatible with plausible views about “what there is.”[13]
He immediately adds that he does not himself believe that the “supposed problem here is one of ontology,” even though it “will be helpful to consider the matter first in this ontological form”. He then introduces Quine’s criterion of “ontological commitment” – but is noticeably cautious in his use of that phrase. Plainly he has some reservation, perhaps about over-simple ontologising.
In considering Quine’s criterion it will be useful to distinguish two theses. The first is that if you refer to an object a, then for some x, x = a and you are referring to x. This is simply an instance of the principle often called existential generalisation – but not to beg questions, I’ll follow Graham Priest[14] and refer to the quantifier as the particular quantifier, and the principle as particular generalisation. The second thesis is that any object on which one can quantify exists. Call this the quantificational criterion of existence. Putting the quantificational criterion together with particular generalisation then yields an ancient idea about existence, namely, that whatever we can refer to in thought or discourse exists. Since particular generalisation is a logical truism – if you refer to an object you refer to something – the only way to resist this conclusion is to reject the quantificational criterion of existence.
There is another ancient idea about existence: that to exist is to have causal status or standing. Call this the causal criterion of existence. Let me stress that in itself this criterion says nothing about the nature of causation; it is compatible with a wide range of views about what causation and causal dependence is. As is familiar, however, trying to combine these two criteria gives rise to difficulties, as much in meta-ethics as in the philosophy of mathematics. Neither reason relations nor sets seem to have any kind of causal standing, on any reasonable interpretation of causal standing.
Scanlon is committed to rejecting the causal criterion as a general condition on existence, for he denies that there is any such general condition: “the only thing common to existential claims across all domains is the purely formal logic of the existential quantifier.” In contrast, he accepts the quantificational criterion – but he accepts it as a criterion for existence within a domain. Thus he adds that

our ontological commitments in this general sense [the sense captured by the quantificational criterion] do not represent a claim on our part about what the world contains, in any meaningful sense of “the world.”[15]

To ask what there ‘really is’, ‘in the world as such,’ is, on his view, to ask a question about an empty notion. Furthermore, forming a general disjunctive concept of existence across domains is he thinks a pointless manoeuvre. How seriously ‘ontological’, then – if I may put it this way – is his acceptance of the quantificational criterion?

§5 To examine the question we can ask whether there is a domain of fictions, in Scanlon’s sense of ‘domain’. The “purely formal logic” of the quantifier applies: we can consider, for example, the proposition that there is no greater fictional detective than Sherlock Homes, and infer that if it is true then Sherlock Holmes is at least as great as Hercule Poirot. We can ask how many fictional detectives there have been. We thus seem to be committed, by Scanlon’s criteria, to the existence of fictional objects – of course in the fictional domain. For, if we are inclined to try to paraphrase away these apparent references to fictional detectives, that implies that we do, after all, have in mind some trans-domain condition on existence other than, and richer than, the quantificational criterion – perhaps the causal criterion. If the quantificational criterion is all that is in play, there is no ground for questioning our “ontological commitment” to fictional objects.
Yet we certainly say things like
(1) There are many characters in War and Peace, some of whom existed and some of whom did not.
In this sentence it seems prima facie obvious that ‘exist’ or ‘really exist’ is used as an ontological predicate, while the verb ‘to be’ is used in an ontologically non-committal sense. But according to the quantificational criterion, if we quantify over characters in War and Peace, we are committed to their existence. So if we take the example at face value, we should reject the quantificational criterion.
In the example, I’ve used the verb ‘to be’ to express the particular quantifier. In ordinary discourse, we can use both the verbs ‘to exist’ and ‘to be’ in that quantifier role. It would, for example, be perfectly intelligible to say
(2) There exist characters in War and Peace who are not real and characters who are
Or we can emphasise, with ‘really exist:’
(3) Some of the characters in War and Peace really existed and some did not.
I don’t think that common sense has any difficulty in distinguishing between ontologically commital, and ontologically non-commital, quantification, despite this flexibility of language. Nevertheless, the quantificational criterion of existence may get some illegitimate plausibility from the fact that the word ‘exists’ can be used both to express the particular quantifier and in the ontological predicate role. So in what follows I will use ‘exist,’ and only ‘exist’, as the ontological predicate, and I will not use it at all in the quantifier role.
If we can refer to the non-existent, an ontological predicate is needed. If I assert that phlogiston does not exist, I seem to refer to phlogiston. Then by particular generalisation, I refer to some x such that x = phlogiston and x does not exist.
Now Scanlon does not allow for a non-quantificational, irreducibly predicative use of ‘exists’. He takes ‘exists’ to express a quantifier, though of course a domain-centred quantifier. Presumably he might say that the War and Peace sentence means that some x in the domain of fictions is a character in that novel and some x in the domain of causes is a character in that novel. What then about the statement that phlogiston does not exist? Presumably this says that some x in the domain of posits is phlogiston but no x in the domain of causes is phlogiston.
Despite the element that this analysis has in common with Quine’s criterion of ontological commitment, it seems to me that it would be best presented as rejecting Quine’s criterion – in the sense of rejecting it as empty: as purporting to say something where there is nothing to say. Such a view might be put in this way. We know how to establish truths containing quantifiers in their various domains. There are fictional detectives in the domain of fictions, electrons in the domain of causes, and reason-relations in the normative domain. All these domains of discourse are legitimate. To talk about these truths as revealing our “ontological commitments” in various domains seems to say something more, but in fact does no more than repeat, in a misleading way, that these truths are truths.
Scanlon rejects ontology when it is understood as a “domain-transcending idea like Quine’s idea of ontological commitment.” “I am rejecting [he says] this general idea of existence and arguing that genuine ontological questions are all domain-specific.”[16] But, he adds,
To say this is not to deny that there are important and interesting metaphysical or ontological questions. It is only to say that these questions are domain-specific.[17]
However the questions Scanlon has in mind are not I think metaphysical questions. Take the “genuine ontological questions” first. You can ask whether Higgs particles really exist, or whether a fictional detective greater than Sherlock Holmes exists. The first question is ontological; you might doubt whether the second is – However, given Scanlon’s domain-centred view of existence it’s not clear that he should doubt it. Nevertheless I think he would agree that neither question is a metaphysical question. If they can be settled they are settled, respectively, by the methods of physics and of literary discussion of detective fiction. The same applies if you ask whether reason relations really exist. The answer is that they do, if there are reasons. And it is first-order normative discussion that settles that there are reasons. This, as I say, is something with which Scanlon would agree.
The position then is that there is no distinctively metaphysical question to ask about existence or about what exists. This is reminiscent of Carnap’s views about internal and external questions, as Scanlon notes.[18] What about the other questions that Scanlon has in mind as domain-specific ‘metaphysical questions’? They seem to me to be questions about the logical syntax of the domain – about how best to regiment its discourse. So for example the question whether we should distinguish facts that are reasons from other facts which constitute them as reasons seems to me to be a question about what is the most useful way to regiment discourse about reasons – the answer may vary depending on the inquiry. I do not find it useful to call such questions metaphysical, but I accept that this truly is a terminological matter.

§6 Scanlon’s view is elegantly minimal. But it does not seem to me to capture everything we think. To me it seems plausible that the War and Peace example, as in (1), deploys a predicative use of ‘exist’ which is (i) ontological in a domain-transcendent sense and (ii) privileges the domain of causes: to exist in ‘real life’, ‘to really exist’, is to have causal standing. That is what Napoleon has and Pierre does not have. There is such a thing as the world; it is the domain of causes and Napoleon is in it while Pierre is not. Nonetheless there are truths about Pierre, just as there are truths about Napoleon. Since we can refer to the non-existent, references to Pierre do not need to be paraphrased out of these truths for them to be seen as truths.
This is cognitivist irrealism about fictional objects. It rejects the quantificational criterion of existence. And I’ll add that it accepts the causal criterion of existence, although this further thesis calls for more discussion than I can give it here.

§7 The question we can now raise is whether a similar irrealist cognitivism is appropriate for normative objects – specifically, for reason relations. I believe it is. But I should say straightaway that that does not entail that reason relations are fictional objects, or non-existent theoretical posits like phlogiston. To take the first view is to be a fictionalist, to take the second is to be an error theorist. It important to see that these are not the only options. The view I want to defend is in fact much closer to Scanlon’s than to these, despite the differences being discussed in this paper.
The key point is that statements about fictional objects and putative but non-existent reals have mind-dependent truth conditions, whereas statements about reason relations do not. But before developing it, I need to make a point about the causal criterion of existence.
The causal criterion (§4) is important in various contexts. In particular, however, it has an implication for knowledge. It entails that it is perfectly legitimate to ask, of any known mind-independent existent, by what mode of receptivity, in Kant’s term, one knows of it. It is legitimate because, since existence is causal standing, knowledge of objects that exist independently of one’s mind must at some point involve a causal relation – a causal transmission terminating in a receptor located in oneself as knower. As I’ve already said, in using the notion of cause here I am not adjudicating among views of what may be involved in A producing B, and I’m not, in particular, assuming that whatever it is must be naturalistically reducible. Even so the essential point remains: knowledge of mind-independent existents must involve some form of receptivity, however that receptivity operates.
Suppose then that we are persuaded, for example by Scanlon’s discussion of the epistemology of the normative, that knowledge of reason-relations or of sets, which he also discusses, involves no such receptivity – just spontaneity in the knowers, to use the term Kant contrasts to receptivity. Then we must conclude that reason-relations and sets are not mind-independent existents. This conclusion might tempt one to think that they are mind-dependent existents. But it also leaves open the possibility that reason-relations and sets are mind-independent inexistents. Unlike fictional objects, which are mind-dependent objects of imagination, and non-existent theoretical posits, which are mind-dependent objects of cognition, reason relations are mind-independent objects of cognition.
The only condition that can legitimately be placed on reference is that we should know, and be able to communicate to each other, what we are talking about – what our topic is. That condition requires that we should be able to anchor our reference in some inter-subjectively intelligible way. When what we are talking about is Sherlock Holmes the condition is amply satisfied. He’s the detective Conan Doyle invented, the one said to live in Baker Street, and so on. As for phlogiston, it’s a theoretical entity postulated by 18th century scientists to explain phenomena which we now think of as oxidisation.
Contrary to Parmenidean intuitions, presentation as a de re object of thought and discourse does not in itself entail existence. It’s true of Sherlock Homes that you and I are talking about him, even though he does not exist.  Of course, since he does not exist he does not have any properties that presuppose existence, but being talked about is not – unlike being a detective, living in Baker Street, or playing the violin – an existence-presupposing property, any more than being a fictional detective, or being the greatest fictional detective, is an existence-presupposing property.

§8 So let us turn to reason-relations and sets. As with Sherlock Holmes and phlogiston, the fact that they are legitimate objects of reference and quantification, and that there are truths about them, does not entail that they exist. Crucially however we must not to jump from this to the conclusion that they are fictional objects like Sherlock Homes, or, like phlogiston, putative existents that turn out not to exist. These two kinds of non-existent referent – fictions and putative existents – are imagined or invented in the one case, postulated or hypothesised in the other. Only in virtue of these mental acts do they become possible objects of reference; if the chain of reference to them is traced back it must arrive at some mental acts of this kind. The possibility of referring to them, and therefore of making true or false statements about them, presupposes the occurrence of such mental acts. That is not the case either with sets or with reason relations. The possibility of referring to the set of planets, for example, does not presuppose that anyone has imagined or invented, postulated or hypothesised, such a set. That set is a possible object of reference that is mind-independent. Likewise with reason relations: for example the possibility of referring to the reason relation that holds between the fact that an action will be pleasant, and that action, does not presuppose that anyone has imagined, invented or postulated such a reason relation.
The possibility of referring to this reason relation does of course depend – as always – on the possibility of anchoring the reference, with sufficient adequacy for us to know what we are talking about. In the case of reason relations that is done by semantic ascent. We can say, for example, that the fact that some action will be pleasant is always a reason to do it. Ascending an order, we can equally say that the fact that it will be pleasant always stands in a reason relation to doing it. We might say, for example, that it always stands in a pro tanto reason relation to doing it but not always in a sufficient reason relation. When we say these things we refer to reason relations. And since the first-order statement is objectively true or false – that is, its truth value does not depend on what we think or feel or have decided as to its truth value – it trivially follows that the second-order statement is in exactly the same sense objectively true or false.
The second-order statement quantifies over reason-relations. It says that some reason relation holds between the fact that doing something will be pleasant and doing it. From a purely logico-semantic point of view that is fine. Semantics assigns semantic values; it will assign values to reason predicates and to singular terms denoting reason relations. To get from that to the conclusion that reason relations exist we need the quantificational criterion of existence.  But that is a metaphysical not a semantic thesis. It goes beyond semantics, in a way that questions about whether the Higgs particle exists do not go beyond physics. Physics purports to deal with what exists. That is its mission. Hence hypothesising a particle is making a postulate about what exists. Semantics, in contrast, purports to deal with what we talk about when we use various bits of language. That is its mission. The further idea that whatever we talk about exists is entirely extraneous to it.
Irrealism about reason relations rejects that idea. It holds to a different metaphysical thesis, namely that the real, the existent, is that which has causal standing. Reason relations, then, are irreal, inexistent, but objective, mind-independent. I could sum this up by way of a well-known remark often attributed to Kreisel, and which I know Scanlon likes: what matters for mathematics is the objectivity of mathematical statements, not the existence of mathematical objects. Likewise for normativity: what matters for normativity is the objectivity of normative statements, not the existence of normative objects. Normative statements are objective: normative objects are irreal. They are, nonetheless, possible objects of reference.
Another way to illuminate the irrealist position is to note how it brings us back to what is nowadays often seen as an old-fashioned and question-begging distinction: the distinction between factual and normative propositions.
The distinction is unpopular because it clashes with a popular idea: factualism – the thesis that all propositional content is factual content. A correspondence theory of truth is committed to factualism, since on this view what determines the content of an assertion is its truth maker, and truth makers are facts. In another way a person who holds that the only tenable notion of a fact is the Fregean notion, according to which a fact is just a truth, is also committed to it. So it may be that Scanlon is committed to it.
There is however another, and very common, notion of fact. On this understanding a fact is the instantiation of some real property or relation, that is, of a property or relation that has causal standing. This, for example, is the notion of fact at stake when we talk, as we often do, about causal relations between facts (the fact that the crane stays upright is due to the fact that sufficient counter-weights have been installed). It is obviously a truism that if you assert that some fact obtains, in this sense of fact, your statement is true just if that fact does obtain. Since this is a truism, it must be respected by any account of truth; it should therefore be clearly distinguished from the correspondence account. It does not, unlike the correspondence account, entail that every assertion is factual or that all content is factual content. And indeed on the irrealist view a purely normative proposition about reason relations has no factual content in this common sense of fact, since reason relations are irreal. A purely normative proposition says something alright; it just doesn’t say anything factual. Thus to assert that the fact that some action will be pleasant is always a reason to do it says something normative about that fact – it doesn’t say anything factual about that fact.
This standpoint, I believe, captures our pre-theoretical view about normativity more closely and clearly than the standpoint proposed by Scanlon, which introduces a domain-centred view of existence while still applying the quantificational criterion. But, someone may ask, isn’t this standpoint actually a mere terminological variant of Scanlon’s view? I see the point, but my tentative inclination is to say no. True, it is a matter of convention how we use the word ‘exist’. And that word occurs in ordinary discourse in a variety of ways (as does the word ‘fact’). It sometimes plays the quantificational role and it sometimes plays the ontological role. So my version of cognitivism does some regimenting of ordinary talk just as Scanlon’s does. Moreover, if Scanlon is willing to apply his domain-centred notion of existence to fictional objects then the two accounts are indeed very close. Both deny that simple cognitivism about the normative requires any form of metaphysical realism, natural or non-natural. Still, I think my account better clarifies why that is so, through a better account of ontic criteria, and for that reason I don’t think the two accounts are simply terminological variants. Unlike Scanlon, I think we do have a conception of the world, or reality; unlike the realist, I don’t believe we think it contains ‘normative facts.’ Reality is the unified causal order through which we find our way. We are in it, in a way in which reason relations are not.

My thanks to members of the audience at the Rio-2015 Metaethics Conference who commented on this paper; especially to my respondent, Rodrigo Gouvea.
[1] T. M. Scanlon, Being Realistic About Reasons, Oxford: Oxford University Press, 2014. This is a revised version of the Locke lectures which Scanlon delivered in Oxford in 2009.
[2] p. 1
[3] p. 2
[4] p. 34; see also pp 25-26
[5] I’ve argued that they can all be analysed that way in John Skorupski, The Domain of Reasons, Oxford: Oxford University Press, 2010.
[6] p. 2
[7] p. 2
[8] p. 31
[9] ‘The idea that normative judgements are correct when they correspond to the normative facts is no explanation if these “facts” are, as I have suggested, merely “the reflection of true thoughts”’ p. 66. Cp pp. 24 n. 9, 62.
[10] p. 21
[11] p. 19
[12] p. 25, and 25 n. 12.
[13] p. 17
[14] Priest, Graham, Towards Non-Being: the Logic and Metaphysics of Intentionality, Oxford: Oxford University Press, 2005.
[15] p. 24
[16] p. 24 note 10.
[17] p. 25.
[18] p. 19, note 3, p. 23.



Pièces mécaniques en chocolat - Lucca